CN112161805A - Bearing tiny fault diagnosis method based on time series scale analysis and CNN - Google Patents

Bearing tiny fault diagnosis method based on time series scale analysis and CNN Download PDF

Info

Publication number
CN112161805A
CN112161805A CN202010917461.0A CN202010917461A CN112161805A CN 112161805 A CN112161805 A CN 112161805A CN 202010917461 A CN202010917461 A CN 202010917461A CN 112161805 A CN112161805 A CN 112161805A
Authority
CN
China
Prior art keywords
scale
data
sequence
bearing
time series
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010917461.0A
Other languages
Chinese (zh)
Inventor
李智
雷增卷
李孟超
于萍
胡波
王磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhongke Innovation Beijing Technology Co ltd
China Three Gorges Corp Fujian Branch
Original Assignee
Zhongke Innovation Beijing Technology Co ltd
China Three Gorges Corp Fujian Branch
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhongke Innovation Beijing Technology Co ltd, China Three Gorges Corp Fujian Branch filed Critical Zhongke Innovation Beijing Technology Co ltd
Priority to CN202010917461.0A priority Critical patent/CN112161805A/en
Publication of CN112161805A publication Critical patent/CN112161805A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G01M13/045—Acoustic or vibration analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention discloses a bearing tiny fault diagnosis method based on time scale analysis and a Convolutional Neural Network (CNN). Aiming at the problem that the early tiny fault of the bearing is difficult to distinguish due to noise interference, the invention designs a diagnosis method, which comprises the following steps: 1) collecting bearing data, and eliminating autocorrelation of the sequence through differential operation; 2) calculating a scale curve of the difference sequence by using a de-trending fluctuation analysis algorithm; 3) constructing a training set and a testing set which are composed of scale sequences of different faults; 4) constructing a convolutional neural network framework and initializing network parameters; 5) the network is trained and verified using the test set. According to the invention, the hidden early fault information in the data is analyzed and mined through the time sequence, and the early fault information is diagnosed by using a deep learning method, so that the problem of early fault diagnosis under noise interference can be effectively solved.

Description

Bearing tiny fault diagnosis method based on time series scale analysis and CNN
Technical Field
The invention relates to the field of fault diagnosis of mechanical equipment, and provides a bearing tiny fault diagnosis method based on time series scale analysis and CNN.
Background
The bearing is one of the most important and most prone parts of the rotating machinery, and statistically, about 30% of mechanical faults are caused by the bearing, so that diagnosing the bearing fault is of great significance to the safe and reliable operation of the machinery. The key point of realizing diagnosis is to extract the fault characteristics of the bearing vibration signal, and the signal presents non-stationarity and non-linearity due to the problems of noise, transmission path and the like, so that the fault characteristics are easily interfered and covered, especially the characteristics of early minor faults. And timely identifying the characteristics of the fault at the weak stage has great effect on the maintenance regulation and the service life prediction of the whole equipment.
Scholars at home and abroad carry out a great deal of research work aiming at the identification of the tiny faults of the bearing. The Liu leading peak applies a resonance sparse decomposition method to weak fault feature extraction of the rolling bearing, and is compared with weak fault feature extraction of envelope analysis and wavelet analysis; lukasz Jedlinski et al use a support vector machine and continuous wavelet transformation for early fault diagnosis of a gearbox; the Gongwenfeng and the like provide a new method of the CNNs-SVM for tiny fault diagnosis of the motor bearing.
The extraction of the bearing vibration signal features is the key to realize fault diagnosis. The traditional method such as Fourier transform is only suitable for stationary signals, and can not effectively process non-stationary signals; wavelet analysis lacks adaptivity to the signal under study; empirical mode decomposition has the problems of over-enveloping, under-enveloping, end-point effects and the like. In 1994, PENG et al proposed a new non-stationary signal analysis method, detrended fluctuation analysis, to measure the long-range correlation of time series. The method can effectively filter various irrelevant trends in the non-stationary signals, and realize the classification of different objects according to the scale index. The plum force and the like realize the identification of 5 faults of the gearbox by using a single scale function; liyang et al utilize multi-fractal detrending fluctuation analysis to implement fault diagnosis of wind turbine bearings.
The invention extracts the fault characteristics of the bearing vibration signal by utilizing the unique processing capacity of a trend-free fluctuation analysis method in a non-stationary sequence, and realizes the diagnosis of early tiny faults by combining the strong learning and classification capacity of a convolutional neural network, and the method comprises the following specific steps of: the autocorrelation of the acquired vibration signals is eliminated by using a difference algorithm, the characteristics of the processed sequence are extracted by using a detrending fluctuation analysis method to form a scale curve, the scale curve is converted into a two-dimensional image data set and used as the input of a convolutional neural network, and the classification of the tiny faults is realized by training and learning.
Disclosure of Invention
Aiming at the problem that the bearing early-stage tiny faults are easily interfered or covered by noise and the like, the bearing tiny fault diagnosis method based on time series scale analysis and CNN is provided, the advantages of processing unstable data and extracting signal characteristics by using a trend-free fluctuation analysis algorithm are utilized, and strong learning classification capability of a convolutional neural network is combined, so that the tiny faults are identified.
A bearing tiny fault diagnosis method based on time series scale analysis and CNN is characterized by comprising the following steps:
A. collecting bearing data, and eliminating autocorrelation of the sequence through differential operation;
B. calculating a scale curve of the difference sequence by using a de-trending fluctuation analysis algorithm;
C. constructing a training set and a testing set which are composed of scale sequences of different faults;
D. constructing a convolutional neural network framework and initializing network parameters;
E. the network is trained and verified using the test set.
Preferably, in step a, the method for eliminating sequence autocorrelation is as follows:
assuming a time sequence x (i) of length N, (i ═ 1,2, …, N), a sequence of averages is constructed:
Figure RE-GDA0002805573670000021
Figure RE-GDA0002805573670000022
preferably, in step B, the method for calculating the scale curve of the difference sequence by using the detrending fluctuation analysis algorithm is as follows:
b1 New sequence to remove autocorrelation
Figure RE-GDA0002805573670000023
Is divided into K segments according to length N, where K is [ N/N ]]. Since the length of the sequence is not necessarily integral multiple of the time scale n, in order to make full use of the data and avoid data loss, when dividing the small segments, the small segments are divided once respectively from front to back and from back to front, so as to obtain 2K segments.
B3, fitting the polynomial trend of each section of data by adopting a least square method, and recording the fitting trend as Pk(i),(k＝1,2,…,2K；i＝1,2,…,n)。
Pk(i)＝a0+a1i+a2i2+…+akik
B4, calculating covariance function F (k, n) of each data:
when K is 1,2, …, K,
Figure RE-GDA0002805573670000031
when K is K +1, K +2, …,2K,
Figure RE-GDA0002805573670000032
b5, calculating the mean value of F (K, n) for 2K intervals, and obtaining a fluctuation function F (n):
Figure RE-GDA0002805573670000033
b6, changing the time scale n, repeating B1-B6, finding out the power law relation between F (n) and n, namely drawing an lnF (n) -qlnn image.
Preferably, in step C, the data set includes a set of scale curves V ═ lnf (n), since the convolutional neural network processes the two-dimensional image data, the data set is to be two-dimensional, the window length of the original data is set to be W, time series scale analysis is performed on the data of each window to obtain a scale sequence, each set of scale sequences is taken as a line, a two-dimensional data set V is obtained, and the data set is divided into a training set and a test set according to a certain proportion.
Preferably, in step D, the convolutional neural network has a structure including an input layer, a convolutional layer, a pooling layer, a fully-connected layer, and an output layer. In the invention, an input layer is two-dimensional image data formed by a scale curve, a convolution layer and a pooling layer are respectively provided with two layers, the size of a convolution kernel is 3 multiplied by 3, the moving step length is 1, the filling mode is non-filling, the activation function uses a Relu function, the pooling method adopts average pooling, the output of an output layer is state classification, a Softmax classifier is used, 1000 corresponds to a normal state, 0100 corresponds to an inner ring fault, 0010 corresponds to an outer ring fault, and 0001 corresponds to a rolling body fault.
Preferably, in the step E, a training set is input, network parameters are obtained through forward propagation, an error between an output result and an expected target is judged, if the error is within an allowable range, training is completed, and a test set is input for verification; if the error is not converged, updating the network parameters by using back propagation until the error is converged.
Drawings
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
FIG. 1 is a diagram of an apparatus of an embodiment of the present invention
FIG. 2 is a flowchart of fault diagnosis in the present embodiment
FIG. 3 is a diagram of the fault signals collected in the present embodiment
FIG. 4 is an exemplary graph of a calibration curve for an inner ring failure in the present embodiment
FIG. 5 is a diagram of converting one-dimensional data into two-dimensional image in the present embodiment
FIG. 6 is a basic structure diagram of a convolutional neural network
The embodiment is applied to a rolling bearing experiment platform which is composed of a motor, a torque transmission device and a load motor, and is shown in figure 1. The fault bearing is arranged at the driving end or the fan end to be tested respectively, fault defects are manufactured by an electric spark machining method, and the fault defects are divided into faults of an inner ring, an outer ring and a rolling body according to fault positions. In this case, the drive-end faulty bearing data and the normal bearing data measured at a sampling frequency of 12kHz are selected for analysis. The SKF bearings used in the experiments were set to three failure diameters, 0.007 inches, 0.014 inches and 0.021 inches, respectively, representing different degrees of early failure.
Referring to the diagnostic flow of fig. 2, the specific steps and implementation details are as follows:
A. collecting bearing data, and eliminating autocorrelation of the sequence through differential operation;
B. calculating a scale curve of the difference sequence by using a de-trending fluctuation analysis algorithm;
C. constructing a training set and a testing set which are composed of scale sequences of different faults;
D. constructing a convolutional neural network framework and initializing network parameters;
E. the network is trained and verified using the test set.
In the step A, data of three faults under different fault degrees are respectively collected and recorded as
Figure RE-GDA0002805573670000041
Where i is 1,2,3 represents different fault types, j is 1,2,3 represents different fault degrees, and the collected data is shown in fig. 3. The acquired data needs further processing due to problems of noise interference, strong correlation and the like, and the method for eliminating the sequence autocorrelation comprises the following steps:
for any type of fault signal time series x (i), (i ═ 1,2, …, N), a sequence of averages is constructed:
Figure RE-GDA0002805573670000051
Figure RE-GDA0002805573670000052
in the step B, the method for calculating the scale curve of the difference sequence by using the detrending fluctuation analysis algorithm comprises the following steps:
b1 New sequence to remove autocorrelation
Figure RE-GDA0002805573670000053
Is divided into K segments according to length N, where K is [ N/N ]]. Since the length of the sequence is not necessarily integral multiple of the time scale n, in order to make full use of the data and avoid data loss, when dividing the small segments, the small segments are divided once respectively from front to back and from back to front, so as to obtain 2K segments.
B2, fitting the polynomial trend of each section of data by adopting a least square method, and recording the fitting trend as Pk(i),(k＝1,2,…,2K；i＝1,2,…,n)。
Pk(i)＝a0+a1i+a2i2+…+akik
B3, calculating covariance function F (k, n) of each data:
when K is 1,2, …, K,
Figure RE-GDA0002805573670000054
when K is K +1, K +2, …,2K,
Figure RE-GDA0002805573670000055
b4, calculating the mean value of F (K, n) for 2K intervals, and obtaining a fluctuation function F (n):
Figure RE-GDA0002805573670000056
b5, changing the time scale n, repeating B1-B4, finding out the power law relation between F (n) and n, namely drawing an lnF (n) -qlnn image. By way of example, FIG. 4 shows a scale plot of inner ring failure at 0.007 inches.
In the step C, the data set includes a scale curve forming V ═ lnf (n), because the convolutional neural network processes the two-dimensional image data, the data set is to be two-dimensional, the window length of the original data is set to be W, the data of each window is analyzed by adopting time series scale to obtain a scale sequence, each group of scale sequences is taken as a line, then a two-dimensional data set V is obtained, and the data set is divided into a training set and a test set according to a certain proportion. The transition diagram is shown in fig. 5.
In step D, the structure of the convolutional neural network includes an input layer, a convolutional layer, a pooling layer, a full-link layer, and an output layer, and the basic structure is shown in fig. 6. In the invention, an input layer is two-dimensional image data formed by a scale curve, a convolution layer and a pooling layer are respectively provided with two layers, the size of a convolution kernel is 3 multiplied by 3, the moving step length is 1, the filling mode is non-filling, an activation function uses a Relu function, a pooling method adopts average pooling, the output of an output layer is state classification, a Softmax classifier is used, 1000 corresponds to a normal state, 0100 corresponds to an inner ring fault, 0010 corresponds to an outer ring fault, 0001 corresponds to a rolling element fault, and relevant parameters are set as shown in a table 1.
TABLE 1 CNN parameter settings
Figure RE-GDA0002805573670000061
Step E, inputting a training set, obtaining network parameters through forward propagation, judging the error between an output result and an expected target, finishing training if the error is within an allowable range, and inputting a test set for verification; if the error is not converged, updating the network parameters by using back propagation until the error is converged.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be conceived by those skilled in the art within the technical scope of the present invention will be covered by the scope of the present invention.

Claims (6)

1. A bearing tiny fault diagnosis method based on time series scale analysis and CNN is characterized by comprising the following steps:
A. collecting bearing data, and eliminating autocorrelation of the sequence through differential operation;
B. calculating a scale curve of the difference sequence by using a de-trending fluctuation analysis algorithm;
C. constructing a training set and a testing set which are composed of scale sequences of different faults;
D. constructing a convolutional neural network framework and initializing network parameters;
E. the network is trained and verified using the test set.
2. The bearing minor failure diagnosis method based on time series scale analysis and CNN according to claim 1, characterized in that:
in step a, the method for eliminating sequence autocorrelation is as follows:
assuming a time sequence x (i) of length N, (i ═ 1,2, …, N), a sequence of averages is constructed:
Figure RE-FDA0002805573660000011
Figure RE-FDA0002805573660000012
3. the bearing minor failure diagnosis method based on time series scale analysis and CNN according to claim 1, characterized in that:
in the step B, the method for calculating the scale curve of the difference sequence by using the detrending fluctuation analysis algorithm comprises the following steps:
b1 New sequence to remove autocorrelation
Figure RE-FDA0002805573660000013
Is divided into K segments according to length N, where K is [ N/N ]](ii) a Because the length of the sequence is not necessarily integral multiple of the time scale n, in order to make full use of the data and avoid data loss, when dividing small segments, the small segments are divided from front to back and from back to front respectively once, so that 2K segments are obtained;
b2, fitting the polynomial trend of each section of data by adopting a least square method, and recording the fitting trend as Pk(i),(k＝1,2,…,2K；i＝1,2,…,n)；
Pk(i)＝a0+a1i+a2i2+…+akik
B3, calculating covariance function F (k, n) of each data:
when K is 1,2, …, K,
Figure RE-FDA0002805573660000021
when K is K +1, K +2, …,2K,
Figure RE-FDA0002805573660000022
b4, calculating the mean value of F (K, n) for 2K intervals, and obtaining a fluctuation function F (n):
Figure RE-FDA0002805573660000023
b6, changing the time scale n, repeating B1-B6, finding out the power law relation between F (n) and n, namely drawing an lnF (n) -qlnn image.
4. The bearing minor failure diagnosis method based on time series scale analysis and CNN according to claim 1, characterized in that:
in the step C, the data set includes a scale curve forming V ═ lnf (n), because the convolutional neural network processes the two-dimensional image data, the data set is to be two-dimensional, the window length of the original data is set to be W, the data of each window is analyzed by adopting time series scale to obtain a scale sequence, each group of scale sequences is taken as a line, then a two-dimensional data set V is obtained, and the data set is divided into a training set and a test set according to a certain proportion.
5. The bearing minor failure diagnosis method based on time series scale analysis and CNN according to claim 1, characterized in that:
in the step D, the structure of the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and an output layer; in the invention, an input layer is two-dimensional image data formed by a scale curve, a convolution layer and a pooling layer are respectively provided with two layers, the size of a convolution kernel is 3 multiplied by 3, the moving step length is 1, the filling mode is non-filling, the activation function uses a Relu function, the pooling method adopts average pooling, the output of an output layer is state classification, a Softmax classifier is used, 1000 corresponds to a normal state, 0100 corresponds to an inner ring fault, 0010 corresponds to an outer ring fault, and 0001 corresponds to a rolling body fault.
6. The bearing minor failure diagnosis method based on time series scale analysis and CNN according to claim 1, characterized in that:
step E, inputting a training set, obtaining network parameters through forward propagation, judging the error between an output result and an expected target, finishing training if the error is within an allowable range, and inputting a test set for verification; if the error is not converged, updating the network parameters by using back propagation until the error is converged.
CN202010917461.0A 2020-09-03 2020-09-03 Bearing tiny fault diagnosis method based on time series scale analysis and CNN Pending CN112161805A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010917461.0A CN112161805A (en) 2020-09-03 2020-09-03 Bearing tiny fault diagnosis method based on time series scale analysis and CNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010917461.0A CN112161805A (en) 2020-09-03 2020-09-03 Bearing tiny fault diagnosis method based on time series scale analysis and CNN

Publications (1)

Publication Number Publication Date
CN112161805A true CN112161805A (en) 2021-01-01

Family

ID=73858638

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010917461.0A Pending CN112161805A (en) 2020-09-03 2020-09-03 Bearing tiny fault diagnosis method based on time series scale analysis and CNN

Country Status (1)

Country Link
CN (1) CN112161805A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113740066A (en) * 2021-11-08 2021-12-03 中国空气动力研究与发展中心设备设计与测试技术研究所 Early fault detection method for compressor bearing

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113740066A (en) * 2021-11-08 2021-12-03 中国空气动力研究与发展中心设备设计与测试技术研究所 Early fault detection method for compressor bearing

Similar Documents

Publication Publication Date Title
CN109460618B (en) Rolling bearing residual life online prediction method and system
CN111797567B (en) Bearing fault classification method and system based on deep learning network
Yan et al. An efficient approach to machine health diagnosis based on harmonic wavelet packet transform
CN111046945B (en) Fault type and damage degree diagnosis method based on combined convolutional neural network
CN109782603A (en) The detection method and monitoring system of rotating machinery coupling fault
CN110866365A (en) Mechanical equipment intelligent fault diagnosis method based on partial migration convolutional network
CN112257530B (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN111562108A (en) Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN116226646B (en) Method, system, equipment and medium for predicting health state and residual life of bearing
CN113569990B (en) Strong noise interference environment-oriented performance equipment fault diagnosis model construction method
CN109655266B (en) Wind turbine generator bearing fault diagnosis method based on AVMD and spectrum correlation analysis
CN112378660A (en) Intelligent fault diagnosis method for aero-engine bearing based on data driving
CN113112000B (en) Construction method and application of multilayer convolutional neural network model
CN112729834B (en) Bearing fault diagnosis method, device and system
CN116358871B (en) Rolling bearing weak signal composite fault diagnosis method based on graph rolling network
Chen et al. A visualized classification method via t-distributed stochastic neighbor embedding and various diagnostic parameters for planetary gearbox fault identification from raw mechanical data
CN111076934A (en) Method for diagnosing potential fault of bearing based on S transformation
CN114964782A (en) Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing
CN116089813A (en) Wind turbine generator set transmission chain fault diagnosis method based on machine learning
Liu et al. Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis
CN114861349A (en) Rolling bearing RUL prediction method based on model migration and wiener process
CN112161805A (en) Bearing tiny fault diagnosis method based on time series scale analysis and CNN
CN117892198A (en) Rolling bearing fault diagnosis method and device based on deep learning
CN113283028A (en) Fault diagnosis method for gear of gear box
CN116821762A (en) Mechanical fault diagnosis method based on multi-scale graph attention fusion network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210101